In a high performance computing (HPC) system, the computing performance of an entire system is increased by managing a large number of computing servers and operating them concurrently. In management of a large number of computing servers in this type of HPC system, there may be a case in which control servers collect data such as files are collected from a large number of computing servers.
When a small number of control servers collect files from a large number of computing servers, however, there is a problem that a communication load several times higher than files transmitted from the computing servers is applied to the control servers.
From the viewpoint of mitigating a communication load, there is a method of reducing communication load by compressing a whole or part of data to be transferred. Specifically, in view of a load applied to a central processing unit (CPU) due to compression at the transmission source, the total of the CPU load at the transmission source and the communication load is kept to or below a communication load applied when compression is not performed. In this method, however, the CPU load at the transmission source is considered but a case is not considered in which data is transmitted from a larger number of computing servers to a small number of control servers.
Japanese Laid-open Patent Publication No. 2008-176420 is known as a related art, for example.